Overview
Researching Machine Learning concepts and methods dates back as early as the 1960’s. However, this field experienced a bitter setback shortly after as the networks of that time were not able to learn a simple XOR gate. This lead to a longer period of time where almost no significant papers in the field of Artificial Intelligence (AI) have been published. This is also known as the AI winter.
This has changed both with important contributions from Geoff Hinton and others into this field beginning around the early 2000s as well as dramatic performance improvements of computing hardware in the last years. The most recent years have revealed exiting advances for previously unthinkable problems, which have essentially been solved now by applying Deep Learning methods.
The procedure
Projects typically start with the client having a vague idea that there exists a potential application within their business processes or their business data that can be solved using Machine Learning or even Deep Learning methods. Hence we usually start with a low-scope feasibility study to check whether or not the specific problem can be solved with classical engineering methods or indeed is an interesting application for applying AI methods.
Types of applications
Recently finished projects included applying Deep Learning methods to improve optical character recognition (OCR), audio signal processing on mobile devices as well as people intrusion detection on low resolution, infrared cameras.
Do not hesitate to contact us. We help you identifying potential business problems and data and support you during data training, Model testing and verification as well as deployment and report generation.
Finished Deep Learning projects
-
AI Audio Processing 04.2018 - 12.2019
High performance, concurrent implementation of signal processing and AI algorithms on mobile devices
-
AR/VR SDK 04.2019 - 08.2019
Development of a Machine Learning framework for mobile devices used for AR/VR applications
-
Furniture Recognition 12.2016 - 04.2017
Proof-of-concept for recognizing furnitures based on a few images using Transfer Learning on a pre-trained FFN
-
Vision Algorithms 02.2016 - 12.2017
Development, optimization and parallelization of Computer Vision algorithms for the HALCON machine vision library